refactor(token-meter): make context occupancy durable projection state

Replace the transient `session/model-request` mux frame with ordinary durable
session state. Occupancy now rides two last-wins projection fields instead of a
non-replayable frame that needed removal tombstones and cross-stream fencing.

The frame was the only non-replayable class on the mux stream. Because host and
mux are independent SSE streams with no cross-stream order, a request emitted
before a removal could arrive after `host/session-removed`, and a legitimate
request for a new lifecycle reusing the same id could be fenced by a late
removal. Fixing that needed a lifecycle generation on every frame; the frame
itself was the problem.

Removed: the `session/model-request` frame and schema, the `agent/model-request`
core event, the ApiProxy measurement point, the client-side telemetry map and
removal tombstone, and the synthetic `cancelled` open error used to signal
reconnect through the error channel.

Added: `request/context`, a log-only session event recording the
registration-bound capacity of the route a request resolved to, appended beside
`request/header` from the lookup that already prepared the call and skipped when
the route is unchanged. Capacity stays out of `EpochHeader` because it is
adapter metadata about a route, not an input the request was built from, so it
must not join request reconstruction or header equality.

The `contextPressure` projection pairs the newest provider-reported prompt size
with the newest recorded capacity. The two are deliberately not one atomic
request observation: switching models can pair a fresh capacity with the prior
route's pressure until the next request reports usage. The figure is a
user-facing reference, and this matches how the TUI status line has always
computed occupancy.
This commit is contained in:
Hypatia May
2026-07-30 13:53:08 +08:00
parent fdccc58cef
commit 4819210142
62 changed files with 382 additions and 1362 deletions

View File

@@ -50,7 +50,7 @@ Agent *创建* 由实现 `AgentFactory` 的插件(`dsh-agent-loop`)提供,
生命周期边有两个重要的本地注意事项。`agent/created` 在作用域 setup 之后、会话与 agent 注册表条目都存在之后运行。Setup 是受信任、仅用于组合的代码;紧随其后且不可 veto 的 `agent/session-start` 通知是第一个受支持的启动注入点。`agent/disposed` 始终表示确切 agent 已离开注册表。AgentLoop 在其驱动器静默后发出该事件,而有序 teardown 此时可能仍在分离会话并撤销作用域;直接注册的自定义 agent 自行拥有任何更强的驱动器顺序契约。
大多数拦截点都是协作式 waterfall。轮次作用域的异步 seam 接收一个显式 `AbortSignal`,其中 `signal` 紧邻 waterfall 最终的 `next`;监听器可以配合,但不得将它保留为控制另一轮次的权限。`agent/step` 是派生请求前的串行检查点;由绑定注册项的准备完成调用发起的外层流返回句柄后,失败受收容的 `agent/model-request` 通知会报告其路由,以及从该调用复制的可选上下文容量。它既不是持久状态,也不能证明提供方 I/O 已开始。`agent/request-error` 是失败模型请求的恢复 waterfall:失败步骤关闭后,它接收确切错误、规范化失败事实和信号。拥有恢复权的监听器返回 `{ kind: 'retry' }` 且不调用 `next()`;循环会关闭失败轮次,并打开一个编号重试轮次。`agent/turn-stopping` 在本可完成的轮次关闭前运行。普通排队提示词保持原样。有效的广义取消会先发出只观测的 `agent/cancel-requested` 及其解析后的类型化原因,再清空队列并中止;通知失败会被收容,不能 veto 停止。信号生命周期由[显式取消决策](../../../.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.md)拥有;作用域分发与终止结算由 [agent 作用域 runtime 设计 Agent Note](../../../.agents/notes/implemented/architecture/2026-07-12-agent-scope-runtime-design.md#three-execution-boundaries-are-deliberately-one-way)拥有。
大多数拦截点都是协作式 waterfall。轮次作用域的异步 seam 接收一个显式 `AbortSignal`,其中 `signal` 紧邻 waterfall 最终的 `next`;监听器可以配合,但不得将它保留为控制另一轮次的权限。`agent/step` 是派生请求前的串行检查点,而 `agent/request-error` 是失败模型请求的恢复 waterfall:失败步骤关闭后,它接收确切错误、规范化失败事实和信号。拥有恢复权的监听器返回 `{ kind: 'retry' }` 且不调用 `next()`;循环会关闭失败轮次,并打开一个编号重试轮次。`agent/turn-stopping` 在本可完成的轮次关闭前运行。普通排队提示词保持原样。有效的广义取消会先发出只观测的 `agent/cancel-requested` 及其解析后的类型化原因,再清空队列并中止;通知失败会被收容,不能 veto 停止。信号生命周期由[显式取消决策](../../../.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.md)拥有;作用域分发与终止结算由 [agent 作用域 runtime 设计 Agent Note](../../../.agents/notes/implemented/architecture/2026-07-12-agent-scope-runtime-design.md#three-execution-boundaries-are-deliberately-one-way)拥有。
`PromptDecision.additionalContexts` 是由带标识且冻结的 `UserMessage` 值组成的数组,因此每个上下文都保留自己的标识和来源。获准的提示词与每个附加上下文都会在轮次运行前成为各自独立、面向模型的 `user/message` 事件。包装下游允许决策的监听器会保留其 `content` 与 `additionalContexts`,除非有意替换任一字段;替换获准内容时仍会保留提示词的标识。